Political News Comments Annotated for Hateful, Offensive, and Toxic Content

3,481 social media comments from Twitter, YouTube, and Reddit posted in August 2021. The dataset includes annotations for hateful, offensive, and toxic content, along with researcher-coded rhetorical dimensions.

Use Cases

• Train a classifier to predict hateful, offensive, or toxic labels on social media comments.
• Analyze correlations between rhetorical dimensions coded by researchers and the presence of toxic content.
• Study the prevalence and co-occurrence of hateful and offensive annotations across 3,481 political comments.

Strengths

• Dataset contains 3,481 annotated social media comments.
• Includes multi-platform data from Twitter, YouTube, and Reddit.
• Provides annotations for three distinct content categories: hateful, offensive, and toxic.

Limitations

• Moderate sample size of 3,481 comments may limit statistical power for some analyses.
• Data is from a single month (August 2021), limiting temporal analysis.
• Annotations are from MTurk workers, which may introduce subjective bias.

Provenance

Source

ICPSR Harvested Dataverse

Collection Method

Comments gathered from political news posts and videos on Twitter, YouTube, and Reddit, annotated by MTurk workers and coded by researchers.

Time Range

August 2021

Freshness

Data was collected in August 2021.